SAN and NAS storage

Ceph and software-defined storage

Ceph turns a set of ordinary servers with local disks into a resilient, scale-out storage cluster with block (RBD), file (CephFS) and object (RGW) interfaces. It underpins Proxmox hyper-converged clusters and Kubernetes storage via Rook.

We run Ceph in production and design it conservatively: three or more nodes, separate cluster and public networks, NVMe or SSD for OSDs and no shortcuts on replica counts.

At a glance

InterfacesRBD block, CephFS, RGW S3-compatible object
PlatformsProxmox VE Ceph, Rook on Kubernetes, standalone cephadm
AlternativesLonghorn (Kubernetes block), MinIO (object), GlusterFS (legacy)
Hardware3+ nodes, 10/25 GbE, enterprise SSD/NVMe

What RackLedge does

How we work

Storage projects begin with an honest capacity and performance baseline: how much data, how fast it grows, what IOPS and latency the applications actually need. That baseline decides between an entry SAN, a unified array, a NAS or a hyper-converged design, and it keeps vendors from selling you the biggest box.

We build the network under the storage as carefully as the storage itself: isolated iSCSI paths or dual FC fabrics, multipathing on every host and a failover test before anything goes live. Backups get an immutable copy and a restore test on a schedule.

Migrations are run as projects with a written cutover plan, a rollback path and a validation checklist. Old arrays are decommissioned with certified data destruction and the documentation is handed over.

Related services

More on san and nas storage

Frequently asked questions

How many nodes do I need?

Three minimum for a 3x replica pool that survives a node failure. Five gives you room to do maintenance without running degraded.

Need a hand with this?

Tell us what you are running and what is slowing you down. You get a straight assessment and a plan, with no obligation. Support desk is staffed 24/7.

Get in touch